PATHSenrich: A Web Service Prototype

for Automatic Cultural Heritage Item

Enrichment

Eneko Agirre, Ander Barrena, Kike Fernandez, Esther Miranda,

Arantxa Otegi, and Aitor Soroa

IXA NLP Group, University of the Basque Country UPV/EHU

arantza.otegi@ehu.es

Abstract. Large amounts of cultural heritage material are nowadays

available through online digital library portals. Most of these cultural
items have short descriptions and lack rich contextual information. The
PATHS project has developed experimental enrichment services. As a
proof of concept, this paper presents a web service prototype which allows
independent content providers to enrich cultural heritage items with a
subset of the full functionality: links to related items in the collection
and links to related Wikipedia articles. In the future we plan to provide
more advanced functionality, as available offline for PATHS.

1 Introduction
Large amounts of cultural heritage (CH) material are now available through
online digital library portals, such as Europeana1. Europeana hosts millions of
books, paintings, films, museum objects and archival records that have been digi-
tised throughout Europe. Europeana collects contextual information or metadata
about different types of content, which the users can use for their searches.
The main strength of Europeana lays in the vast number of items it contains.
Sometimes, though, this quantity comes at the cost of a restricted amount of
metadata, with many items having very short descriptions and a lack of rich
contextual information. One of the goals of the PATHS project2 is precisely to
enrich CH items, using a selected subset of Europeana as a testbed[1].
Whithin the project, this enrichment will make possible to create a system
that acts as an interactive personalised tour guide through Europeana collec-
tions, offering suggestions about items to look at and assist in their interpre-
tation by providing relevant contextual information from related items within
Europeana and items from external sources like Wikipedia. Users of such digital
libraries may require information for purposes such as learning and seeking an-
swers to questions. This additional information supports users in fulfilling their
information need, as the evaluation of the first PATHS prototype shows [2].
In this paper we present a web service prototype which allows independent
content providers to enrich CH items. Specifically, the service enriches the items
1
http://www.europeana.eu/portal/
2
http://www.paths-project.eu

T. Aalberg et al. (Eds.): TPDL 2013, LNCS 8092, pp. 462–465, 2013.

c Springer-Verlag Berlin Heidelberg 2013
PATHSenrich: A Web Service Prototype for Automatic CH Item Enrichment 463

with two types of information. On the one hand, the item will be linked to
similar items within the collection. On the other hand, the item will be linked
to Wikipedia articles which are related to it.
There have been many attempts to automatically enrich cultural heritage
metadata. Some projects (for instance, MIMO-DB3 or MERLIN4 ) relate CH
objects with terms of an external authority or vocabulary. Some others (like
MACE5 or YUMA 6 ) adopt a collaborative annotation paradigm for metadata
enrichment. To our knowledge, PATHS is the first project using semantic NLP
processing to link CH items to similar items or external Wikipedia articles.
The current service has limited bandwidth, and provides a selected subset
of the enrichment functionality available internally in the PATHS project. The
quality of the links produce is also slightly lower, although we plan to improve it
in the short future. However, we think that the prototype is useful to demonstrate
the potential to construct a web service for automatically enriching CH items
with high quality information.

2 Demo Description
The web service takes as input one CH item represented following the Europeana
Data Model (EDM) in JSON format, as exported by the Europeana API v2.07 (a
sample record is provided in the interface). The web service returns the following:
– A list of 10 closely related items within the collection.
– A list of Wikipedia pages which are related to the target item.
Figure 1 shows a snapshot of the web service. The service is publicly accessible
following the URL http://ixa2.si.ehu.es/paths_wp2/paths_wp2.pl.
The enrichment is performed by analyzing the metadata associated with the
item, i.e., the title of the item, its description, etc. The next sections briefly
describe how this enrichment is performed.

2.1 Related Items within the Collection

The list of related items is obtained by first creating a query with the content
of the title, subject and description fields (stopwords are removed). The query
is then posted to a SOLR search engine8 . The SOLR search engine accesses an
index created with the subset of Europeana items already enriched offline within
the PATHS project. In that way, the most related Europeana items in the subset
are obtained, and the identifiers of those related items are listed. Note that the
related items used internally in the PATHS project are produced using more
sophisticated methods. Please refer to [1] for further details.
3
http://www.mimo-international.com
4
http://www.ucl.ac.uk/ls/merlin
5
http://www.mace-project.eu
6
http://dme.ait.ac.at/annotation
7
http://preview.europeana.eu/portal/api-introduction.html
8
http://lucene.apache.org/solr/
464 E. Agirre et al.

Fig. 1. Web service interface. It consists of a text area to introduce the input item
in JSON format (top). The “Get EDM JSON example” button can be used to get an
input example. Once a JSON record is typed, click “Process” button to get the output.
The output (bottom) consists on a list of related items and background links.

2.2 Related Wikipedia Articles

For linking the items to Wikipedia articles we follow an implementation similar

to the method described in [3]. This method creates a dictionary, an association
between string mentions with all possible articles the mention can refer to. Our
dictionary is constructed using the title of the Wikipedia article, the redirect
pages, the disambiguation pages and the anchor texts from Wikipedia links.
Mentions are lower-cased and all text between parenthesis is removed. If the
mention links to a disambiguation page, it is associated with all possible articles
the disambiguation page points to. Besides, each association between a mention
and article is scored with the prior probability, estimated as the number of
times that the mention occurs in the anchor text of an article. Note that such
dictionaries can disambiguate any mention, just returning the highest-scoring
article for this particular mention.
Once the dictionary is built, the web service analyzes the title, subject and
description fields of the CH item and matches the longest substring within those
fields with entries in the dictionary. When a match is found, the Wikipedia article
with highest score for this entry is returned. Note that the links to Wikipedia
in the PATHS project are produced using more sophisticated methods. Please
refer to [1] for further details.
PATHSenrich: A Web Service Prototype for Automatic CH Item Enrichment 465

3 Conclusions and Future Work

This paper presents a web service prototype which automatically enriches CH
items with metadata. The web service is inspired in the enrichment work carried
out in the PATHS project, but, contrary to the batch methodology used in the
project, this enrichment is performed online. The prototype has been designed
for demonstration purposes, to showcase the feasibility of providing full-fledged
automatic enrichment.
Our plans for the future include moving the offline enrichment services which
are currently being evaluated in the PATHS project to the web service. In the
case of related Wikipedia articles, we will take into account the context of the
matched entities, which improves the quality of the links [4], and we will include
a filtering algorithm to discard entities that are not relevant. Regarding related
items, we will classify them according to the type of relation [5]. In addition we
plan to automatically organize the items hierarchically, according to a Wikipedia-
based vocabulary [6].

Acknowledgements. The research leading to these results was carried out as

part of the PATHS project (http://www.paths-project.eu) funded by Eu-
ropean Communitys Seventh Framework Programme (FP7/2007- 2013) under
grant agreement no. 270082. The work has been also funded by the Basque
Government (project IBILBIDE, SAIOTEK S-PE12UN089).

References
1. Otegi, A., Agirre, E., Soroa, A., Aletras, N., Chandrinos, C., Fernando, S., Gonzalez-
Agirre, A.: Report accompanying D2.2: Processing and Representation of Content
for Second Prototype. PATHS Project Deliverable (2012),
http://www.paths-project.eu/eng/content/download/2489/18113/version/2/
file/D2.2.Content+Processing-2nd+Prototype-revised.v2.pdf
2. Griffiths, J., Goodale, P., Minelli, S., de Polo, A., Agerri, R., Soroa, A., Hall, M.,
Bergheim, S.R., Chandrinos, K., Chryssochoidis, G., Fernie, K., Usher, T.: D5.1:
Evaluation of the first PATHS prototype. PATHS Project Deliverable (2012),
http://www.paths-project.eu/eng/Resources/
D5.1-Evaluation-of-the-1st-PATHS-Prototype
3. Chang, A.X., Spitkovsky, V.I., Yeh, E., Agirre, E., Manning, C.D.: Stanford-UBC
entity linking at TAC-KBP. In: Proceedings of TAC 2010, Gaithersburg, Maryland,
USA (2010)
4. Han, X., Sun, L.: A Generative Entity-Mention Model for Linking Entities with
Knowledge Base. In: Proceedings of the ACL, Portland, Oregon, USA (2011)
5. Agirre, E., Aletras, N., Gonzalez-Agirre, A., Rigau, G., Stevenson, M.: UBC UOS-
TYPED: Regression for typed-similarity. In: Second Joint Conference on Lexical
and Computational Semantics (*SEM), Atlanta, Georgia, USA (2013)
6. Fernando, S., Hall, M., Agirre, E., Soroa, A., Clough, P., Stevenson, M.: Com-
paring Taxonomies for Organising Collections of Documents. In: Proceedings of
COLING 2012, Mumbai, India (2013)